The ultimate question is whether the distribution of brands in the OPIS data biases our results.
We can make a conjecture based on Hastings’s Table 3, which documents the effects of the
ARCO-Thrifty transaction by “high-share,” “mid-share,” and “low-share” brands.13 (The effect
on ARCO’s premerger stations is estimated separately.) Column 7 of our Table 1 displays the
Hastings classification for each brand. Hastings posits that market share is positively correlated
with the degree of brand loyalty. Consequently, independents compete most closely with lowshare
brands, so that the ARCO-Thrifty transaction should increase prices the most at low-share
branded stations. Indeed, using the W-L data, Hastings finds that prices increased at low-share,
mid-share, and high-share brands by approximately $0.07, $0.05, and $0.03 per gallon, respectively.
Consistent with its low-price strategy, the estimated effect on ARCO resembles that of the
low-share group.